Open Source · AI Tools
Career Launchpad
From outdated LinkedIn to signed offer — coached by Claude.
No repos public yet. The build is underway.
- Status
- Building
- Scaffold live, rubric formalized, dogfooding in progress
- 30%
- Features listed below
- 7
- Tools in the stack
- 6
No. 01 / The problem
What pain does this solve?
The job search stack is a mess of half-working point tools. Jobscan for keywords, Jobright for matching, Final Round for mocks, a dozen browser tabs for application tracking, and a spreadsheet holding it all together. Each tool wants a subscription and owns a slice of your data. Meanwhile LinkedIn's 2026 automation detection catches almost everything, so any tool that tries to auto-apply is a risk to your account. Most candidates either pay for five tools they barely use or run the whole search in a Google Doc and lose track.
No. 02 / The solution
How the tool works
Career Launchpad collapses the entire pipeline into four Claude skills that compose cleanly. The orchestrator detects which phase you need next based on gaps in your state files. LinkedIn Rebuild audits your current profile, extracts keywords from target JDs, and produces a copy-paste rebuild — never automated LinkedIn actions. Role Match Apply polls your target-company watchlist via public ATS JSON, scores every role on Fit (0-100) and ATS (0-100), and tailors a resume and cover letter per application. Interview Prep runs company research, builds a JD-derived question bank, maps STAR stories, and runs mock interviews in three modes. Every write is user-confirmed. Every submission is manual.
- 4-skill chain: orchestrator + LinkedIn rebuild + role match/apply + interview prep
- LinkedIn audit and copy-paste rebuild calibrated to 2026 standards
- Role discovery via public ATS JSON (Greenhouse, Lever, Ashby) with Fit and ATS scoring
- Per-role tailored resume and cover letter with a scorecard and archived JD
- Interview prep with company research, STAR mapping, and three mock modes
- Local-first markdown state — no cloud lock-in, git-friendly, portable
- Three audience modes: Personal, Rare Find candidate, Summit SaaS customer
No. 03 / Technical approach
Architecture decisions
Built as native Claude skills — no SaaS backend, no third-party API keys beyond what Claude already has. All state lives in ~/career-launchpad/ as plain markdown: profile.md, resume-master.md, career-narrative.md, target-companies.md, applications.md, plus per-role folders under tailored/ and interviews/. The role discovery layer queries public Greenhouse, Lever, and Ashby ATS JSON endpoints directly — no scraping, no login. The Fit and ATS scoring rubrics are keyword-matching and requirement-coverage heuristics expressed in prompt instructions, not ML models. Interview mocks use Claude's conversational mode with role-specific personas derived from the question bank. The whole system is designed so the user owns their data and can walk away with a git repo if they ever stop using Claude.
- Claude Code Skills
- Markdown
- Greenhouse ATS
- Lever ATS
- Ashby ATS
- Git
No. 04 / Current status
Where things stand
No. 05 / Building something similar?
Recognize this problem in your own business?
This project started as a real frustration Ryan observed working inside companies. If you're dealing with a similar problem, let's talk. The solution might already be in progress, or we build it together.
Or text or call (248) 882-9055